CamScanner Implementation using OpenCV along with compression¶
In [34]:
import matplotlib.pyplot as plt
import numpy as np
import cv2 as cv
%matplotlib inline
In [35]:
def show_img(img):
fig = plt.figure(figsize=(20,20))
ax = fig.add_subplot(111)
fix_img = cv.cvtColor(img, cv.COLOR_BGR2RGB)
ax.imshow(fix_img)
def show_grey_img(img):
fig = plt.figure(figsize=(20,20))
ax = fig.add_subplot(111)
fix_img = cv.cvtColor(img, cv.COLOR_BGR2GRAY)
ax.imshow(fix_img, cmap ="gray")
In [36]:
img = cv.imread("C:/Users/Jaysurya/Desktop/example_camscanner.PNG")
1. Here's an example of what the ideal output of this project would be:¶
In [37]:
show_img(img)
lets begin with a newspaper image¶
-------------------------------------------------------------------------------------------------------------¶
In [38]:
cross = cv.imread("DATA/crossword.jpg")
show_img(cross)
image converted to a gray scale image¶
In [75]:
cross1 = cv.imread("DATA/crossword.jpg")
show_grey_img(cross1)
In [76]:
cross1.max()
Out[76]:
In [77]:
cross1.mean()
Out[77]:
In [78]:
return1, cross2 =cv.threshold(cross1, 165, 255, cv.THRESH_BINARY)
In [79]:
show_grey_img(cross2)
In [80]:
print(cross2.max()," , ", cross2.min())
In [81]:
fix_img = cv.cvtColor(cross2, cv.COLOR_BGR2GRAY)
#img = plt.imshow(fix_img,cmap="gray")
In [82]:
cv.imwrite("scanned.jpg",cross2)
Out[82]:
Image copyright asignment using blending¶
In [47]:
img1 = cv.cvtColor(cv.imread("DATA/gorilla.jpg"), cv.COLOR_BGR2RGB)
img2 = cv.resize(cv.cvtColor(cv.imread("DATA/watermark_no_copy.png"), cv.COLOR_BGR2RGB),(600,600))
In [48]:
img1.shape
Out[48]:
In [49]:
plt.imshow(img1)
Out[49]:
In [50]:
x_offset = 1925 - 600
y_offset = 1295 - 600
In [51]:
img2.shape
Out[51]:
In [52]:
rows, columns, channel = img2.shape
In [53]:
roi = img1[y_offset:1295,x_offset:1925]
plt.imshow(roi)
Out[53]:
In [54]:
img2_gray = cv.cvtColor(img2,cv.COLOR_RGB2GRAY)
plt.imshow(img2_gray,cmap="gray")
Out[54]:
In [55]:
masked = cv.bitwise_not(img2_gray)
plt.imshow(masked,cmap="gray")
Out[55]:
In [56]:
masked.shape
Out[56]:
In [57]:
white_back = np.full(img2.shape,255,dtype = np.uint8)# have a background for all 3 color channe;s
back_full = cv.bitwise_or(white_back,white_back,mask = masked) # place the masked image onto the white channel background
In [58]:
back_full.shape
Out[58]:
In [59]:
white_back.shape
Out[59]:
In [60]:
plt.imshow(white_back)
Out[60]:
In [61]:
plt.imshow(back_full)
Out[61]:
In [62]:
front = cv.bitwise_or(img2,img2,mask=masked)
plt.imshow(front)
Out[62]:
In [63]:
final_roi = cv.bitwise_or(roi,front)
plt.imshow(final_roi)
Out[63]:
In [64]:
big_img = img1
small_img = final_roi
In [65]:
big_img[y_offset:y_offset+600,x_offset:x_offset+600] = small_img
plt.imshow(big_img)
Out[65]:
